[SymmMem] Add NVSHMEM Fence support to Triton #156474
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/156474
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit ce21940 with merge base ef6d2ce ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
the
oncall: distributed
This was referenced Jun 20, 2025
codingwithsurya
changed the title
codingwithsurya
added
the
release notes: distributed (c10d)
mandroid6
approved these changes
Jun 20, 2025
kwen2501
approved these changes
Jun 20, 2025
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
pytorchmergebot
pushed a commit
that referenced
this pull request
Jun 21, 2025
This PR introduces device-side NVSHMEM completion guarantees via the quiet API in Triton, enabling GPU kernels to ensure all pending remote memory operations are fully complete before proceeding with subsequent operations. Changes: - Added a new `core.extern` wrapper for `nvshmem_quiet` in `nvshmem_triton.py` - Implemented `test_triton_quiet` in `test/distributed/test_nvshmem.py`, including: - A Triton kernel that performs `putmem_block` followed by `quiet()` to ensure completion - Flag-based signaling only after `quiet()` completes, guaranteeing data delivery - Consumer validation that when the completion flag arrives, all data transfers are guaranteed complete Tests: `$ TORCH_SYMMMEM=NVSHMEM python test/distributed/test_nvshmem.py -k test_triton_quiet` Pull Request resolved: #156475 Approved by: https://github.com/kwen2501 ghstack dependencies: #156472, #156473, #156474
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR introduces device-side NVSHMEM memory ordering via the fence API in Triton, enabling GPU kernels to enforce completion and ordering of remote memory operations before subsequent operations proceed.
Changes:
core.externwrapper fornvshmem_fenceinnvshmem_triton.pytest_triton_fenceintest/distributed/test_nvshmem.py, including:putmem_blockoperations separated byfence()callsTests:
$ TORCH_SYMMMEM=NVSHMEM python test/distributed/test_nvshmem.py -k test_triton_fenceStack from ghstack (oldest at bottom):
cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k